Persona-ASR / config.json
RakhatM's picture
Add Persona-ASR backbone + presence-gate checkpoints and model card
b4a7ad7 verified
Raw
History Blame Contribute Delete
1.81 kB
{
"smoke_test": false,
"data_root": "/workspace/LibriMix/storage_dir",
"persona_root": "/workspace/new/data/LibriMix/persona_asr/official_libri3mix_clean_max_16k",
"train_manifests": [
"en=/workspace/new/data/LibriMix/persona_asr/official_libri3mix_clean_max_16k/train_clean_100_official_libri3mix_clean_max_16k_posneg.json",
"kk=/workspace/new/data/LibriMix/persona_asr/kazakh3mix_clean_max_16k_lufsfix/train_clean_100_kazakh3mix_lufsfix_posneg.json"
],
"valid_manifests": [
"en=/workspace/new/data/LibriMix/persona_asr/official_libri3mix_clean_max_16k/val_official_libri3mix_clean_max_16k_posneg.json",
"kk=/workspace/new/data/LibriMix/persona_asr/kazakh3mix_clean_max_16k_lufsfix/val_kazakh3mix_lufsfix_posneg.json"
],
"test_manifests": null,
"run_dir": "/workspace/new/runs/maeda_twohead_lufsfix_fromscratch_20260620",
"wavlm_name": "microsoft/wavlm-base-plus",
"epochs": 100,
"batch_size": 28,
"lr": 0.0001,
"lambda_asr": 0.7,
"lambda_vad": 0.3,
"neg_vad_weight": 1.0,
"warmup_steps": 5000,
"freeze_wavlm_steps": 10000,
"max_seconds": 20.0,
"enrollment_seconds": 5.0,
"vad_threshold": 1e-05,
"num_workers": 6,
"train_limit": null,
"valid_limit": null,
"seed": 7,
"resume_from": null,
"pretrained_backbone": null,
"finetune_from": null,
"include_negatives": false,
"device": "cuda",
"disable_grad_scaler": false,
"en_vocab_size": 30,
"kk_vocab_size": 53,
"kk_chars_sample": [
"<pad>",
"<unk>",
" ",
"0",
"1",
"2",
"3",
"4",
"5",
"6",
"7",
"9",
"\u0406",
"\u0410",
"\u0411",
"\u0412",
"\u0413",
"\u0414",
"\u0415",
"\u0416"
],
"total_steps": 124600,
"world_size": 2,
"architecture": "TSASRADTwoHead (WavLM+FiLM+EN_CTC+KK_CTC+VAD)"
}